Use WarningException, expand CommandResultBuilder

[amorton]

Fixes #1497

Changes the warnings associated commands to use the new ApiExcpetions,
and warnings in the return status now returns a list of error object V2.

To do that needed to improve the way CommandResult was built, so it always
had a status map so the CommandProcessor could append a warning. To do that
expanded the CommandResultBuilder added for the OperationAttempts, removed
the many overloads used for CommandResult, and updated all creation of the
CommandResult to use the builder.

See also #1518 to continue this

What this PR does:

Which issue(s) this PR fixes:
Fixes #

Checklist

• Changes manually tested
• Automated Tests added/updated
• Documentation added/updated
• CLA Signed: DataStax CLA

[tatu-at-datastax]

I have a philosophical objection here: Exceptions should be used for Exceptional situations, failures, and not as internal communication mechanism (i.e. not for convenience). And warnings by their nature are not such things.
It seems to me usage here is due to convenience of using message construction/templating part, and not because Exceptions would make sense otherwise.

Beyond philosophical part, Exceptions are also rather expensive things to use wrt performance: they are inefficient things to use if (although only if) created fresh, with stack traces.

So I am not sure we should take this approach for generating warnings as part of response payload.

[amorton]

Beyond philosophical part, Exceptions are also rather expensive things to use wrt performance: they are inefficient things to use if (although only if) created fresh, with stack traces.

I can make the WarningException overirde fillInStackTrace() so it does not create a stack trace, and I can see a way to run the template without needing to instantiate a an ApiException sub class, I'll spin up a ticket when this is ready. It is already almost there: running the template creates a ErrorInstance record, which is used to construct the ApiException, we can use that and just then need a way to turn them into the CommandErrorV2 for serialisation. Should be a reasonable fast follow with no external impact .

It seems to me usage here is due to convenience of using message construction/templating part, and not because Exceptions would make sense otherwise.

Yup I was optimising for managing the content in the yaml file, and having a standard structure for errors and warnings.

[amorton]

@tatu-at-datastax created this to track the change to not use a full exception #1549